Abstract by Seth Lam
Movie recommendation using matrix factorization
Recommending movies to a user can be a difficult task, since a lot of factors, such as acting, plot, and genre, of a movie can have a significant impact on a movie being (dis)liked by the user. Considering just one of these elements may not be sufficient in predicting which movie a user (dis)likes. We propose to use matrix factorization (MF) to predict movie ratings for a user. We first present at a high level what MF is, which is followed by comparing MF with the user-based and item-based collaborative filtering (CF) approaches, two of the most popular recommendation methods. While most of CF approaches struggle with sparse data, MF overcomes the problem by using unsupervised learning methods, such as dimensional reduction. In addition, we show that MF is the most promising model out of all three in terms of accurately predicting ratings of movies for users using the MovieLen dataset.